Adaptions for Automotive Radar Based Occupancy Gridmaps

Environment models are necessary for autonomous driving. The distinction between drivable and non-drivable underground is elementary. This paper presents adaptions for radar based occupancy gridmaps, which are a common representation of the environment. In contrast to standard occupancy gridmaps or in general standard inverse radar sensor models, our approach works with velocity dependent parameters and extends free space calculations. Consequently, the map quality varies less and the information content of the ego vehicle's immediate vicinity is higher. Experiments with ground truth data show that the proposed algorithm produces accurate environment models in urban scenes.